4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2010, 2014, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/lustre/lustre_user.h
38 * Lustre public user-space interface definitions.
41 #ifndef _LUSTRE_USER_H
42 #define _LUSTRE_USER_H
44 /** \defgroup lustreuser lustreuser
49 #include <libcfs/types.h>
52 # include <linux/quota.h>
53 # include <linux/string.h> /* snprintf() */
54 # include <linux/version.h>
55 #else /* !__KERNEL__ */
56 # define NEED_QUOTA_DEFS
57 # include <stdio.h> /* snprintf() */
59 # include <sys/quota.h>
60 # include <sys/stat.h>
61 #endif /* __KERNEL__ */
62 #include <lustre/ll_fiemap.h>
64 #if defined(__x86_64__) || defined(__ia64__) || defined(__ppc64__) || \
65 defined(__craynv) || defined(__mips64__) || defined(__powerpc64__)
66 typedef struct stat lstat_t;
67 # define lstat_f lstat
68 # define fstat_f fstat
69 # define fstatat_f fstatat
70 # define HAVE_LOV_USER_MDS_DATA
71 #elif defined(__USE_LARGEFILE64) || defined(__KERNEL__)
72 typedef struct stat64 lstat_t;
73 # define lstat_f lstat64
74 # define fstat_f fstat64
75 # define fstatat_f fstatat64
76 # define HAVE_LOV_USER_MDS_DATA
79 #define LUSTRE_EOF 0xffffffffffffffffULL
82 #define LL_SUPER_MAGIC 0x0BD00BD0
84 #ifndef FSFILT_IOC_GETFLAGS
85 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
86 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
87 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
88 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
89 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
90 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
93 /* FIEMAP flags supported by Lustre */
94 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
96 enum obd_statfs_state {
97 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
98 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
99 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
100 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
101 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
115 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
116 __u32 os_fprecreated; /* objs available now to the caller */
117 /* used in QoS code to find preferred
132 * FID is a cluster-wide unique identifier of a file or an object (stripe).
133 * FIDs are never reused.
137 * FID sequence. Sequence is a unit of migration: all files (objects)
138 * with FIDs from a given sequence are stored on the same server.
139 * Lustre should support 2^64 objects, so even if each sequence
140 * has only a single object we can still enumerate 2^64 objects.
143 /* FID number within sequence. */
146 * FID version, used to distinguish different versions (in the sense
147 * of snapshots, etc.) of the same file system object. Not currently
153 static inline bool fid_is_zero(const struct lu_fid *fid)
155 return fid->f_seq == 0 && fid->f_oid == 0;
158 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
159 * MDT-object's FID::f_ver, instead it is the OST-object index in its
160 * parent MDT-object's layout EA. */
161 #define f_stripe_idx f_ver
164 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
167 /* keep this one for compatibility */
168 struct filter_fid_old {
169 struct lu_fid ff_parent;
174 /* Userspace should treat lu_fid as opaque, and only use the following methods
175 * to print or parse them. Other functions (e.g. compare, swab) could be moved
176 * here from lustre_idl.h if needed. */
177 typedef struct lu_fid lustre_fid;
180 * Following struct for object attributes, that will be kept inode's EA.
181 * Introduced in 2.0 release (please see b15993, for details)
182 * Added to all objects since Lustre 2.4 as contains self FID
184 struct lustre_mdt_attrs {
186 * Bitfield for supported data in this structure. From enum lma_compat.
187 * lma_self_fid and lma_flags are always available.
191 * Per-file incompat feature list. Lustre version should support all
192 * flags set in this field. The supported feature mask is available in
196 /** FID of this inode */
197 struct lu_fid lma_self_fid;
201 * Prior to 2.4, the LMA structure also included SOM attributes which has since
202 * been moved to a dedicated xattr
203 * lma_flags was also removed because of lma_compat/incompat fields.
205 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
208 * OST object IDentifier.
216 struct lu_fid oi_fid;
220 #define DOSTID LPX64":"LPU64
221 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
223 struct ll_futimes_3 {
225 __u64 lfu_atime_nsec;
227 __u64 lfu_mtime_nsec;
229 __u64 lfu_ctime_nsec;
233 * The ioctl naming rules:
234 * LL_* - works on the currently opened filehandle instead of parent dir
235 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
236 * *_MDC_* - gets/sets data related to MDC
237 * *_LOV_* - gets/sets data related to OSC/LOV
238 * *FILE* - called on parent dir and passes in a filename
239 * *STRIPE* - set/get lov_user_md
240 * *INFO - set/get lov_user_mds_data
242 /* lustre_ioctl.h 101-150 */
243 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
244 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
245 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
246 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
247 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
248 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
249 /* LL_IOC_RECREATE_OBJ 157 obsolete */
250 /* LL_IOC_RECREATE_FID 157 obsolete */
251 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
252 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
253 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
254 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
255 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
256 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
257 /* IOC_LOV_GETINFO 165 obsolete */
258 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
259 #define LL_IOC_RMTACL _IOW ('f', 167, long)
260 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
261 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
262 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
263 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
264 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
265 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
266 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
267 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
268 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
269 /* lustre_ioctl.h 177-210 */
270 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
271 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
272 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
273 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
274 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
275 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
276 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
277 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
278 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
279 struct lustre_swap_layouts)
280 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
281 struct hsm_current_action)
282 /* lustre_ioctl.h 221-232 */
283 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
284 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
285 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
286 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
287 #define LL_IOC_GET_LEASE _IO('f', 244)
288 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
289 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
290 #define LL_IOC_MIGRATE _IOR('f', 247, int)
291 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
292 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
294 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
296 LL_LEASE_RDLCK = 0x1,
297 LL_LEASE_WRLCK = 0x2,
298 LL_LEASE_UNLCK = 0x4,
301 #define LL_STATFS_LMV 1
302 #define LL_STATFS_LOV 2
303 #define LL_STATFS_NODELAY 4
305 #define IOC_MDC_TYPE 'i'
306 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
307 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
308 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
309 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
311 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
313 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
314 * files, but are unlikely to be used in practice and are not harmful if
315 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
316 * devices and are safe for use on new files. See LU-4209. */
317 /* To be compatible with old statically linked binary we keep the check for
318 * the older 0100000000 flag. This is already removed upstream. LU-812. */
319 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
320 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
321 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
322 O_LOV_DELAY_CREATE_MASK)
324 #define LL_FILE_IGNORE_LOCK 0x00000001
325 #define LL_FILE_GROUP_LOCKED 0x00000002
326 #define LL_FILE_READAHEA 0x00000004
327 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
328 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
329 #define LL_FILE_RMTACL 0x00000020
331 #define LOV_USER_MAGIC_V1 0x0BD10BD0
332 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
333 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
334 #define LOV_USER_MAGIC_V3 0x0BD30BD0
335 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
336 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
338 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
340 #define LOV_PATTERN_RAID0 0x001
341 #define LOV_PATTERN_RAID1 0x002
342 #define LOV_PATTERN_FIRST 0x100
343 #define LOV_PATTERN_CMOBD 0x200
345 #define LOV_PATTERN_F_MASK 0xffff0000
346 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
347 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
349 #define LOV_MAXPOOLNAME 15
350 #define LOV_POOLNAMEF "%.15s"
352 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
353 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
354 #define LOV_MAX_STRIPE_COUNT_OLD 160
355 /* This calculation is crafted so that input of 4096 will result in 160
356 * which in turn is equal to old maximal stripe count.
357 * XXX: In fact this is too simpified for now, what it also need is to get
358 * ea_type argument to clearly know how much space each stripe consumes.
360 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
361 * allocation that is sufficient for the current generation of systems.
363 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
364 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
365 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
366 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
368 #define XATTR_LUSTRE_PREFIX "lustre."
369 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
371 #define lov_user_ost_data lov_user_ost_data_v1
372 struct lov_user_ost_data_v1 { /* per-stripe data structure */
373 struct ost_id l_ost_oi; /* OST object ID */
374 __u32 l_ost_gen; /* generation of this OST index */
375 __u32 l_ost_idx; /* OST index in LOV */
376 } __attribute__((packed));
378 #define lov_user_md lov_user_md_v1
379 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
380 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
381 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
382 struct ost_id lmm_oi; /* LOV object ID */
383 __u32 lmm_stripe_size; /* size of stripe in bytes */
384 __u16 lmm_stripe_count; /* num stripes in use for this object */
386 __u16 lmm_stripe_offset; /* starting stripe offset in
387 * lmm_objects, use when writing */
388 __u16 lmm_layout_gen; /* layout generation number
389 * used when reading */
391 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
392 } __attribute__((packed, __may_alias__));
394 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
395 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
396 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
397 struct ost_id lmm_oi; /* LOV object ID */
398 __u32 lmm_stripe_size; /* size of stripe in bytes */
399 __u16 lmm_stripe_count; /* num stripes in use for this object */
401 __u16 lmm_stripe_offset; /* starting stripe offset in
402 * lmm_objects, use when writing */
403 __u16 lmm_layout_gen; /* layout generation number
404 * used when reading */
406 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
407 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
408 } __attribute__((packed));
410 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
412 if (lmm_magic == LOV_USER_MAGIC_V1)
413 return sizeof(struct lov_user_md_v1) +
414 stripes * sizeof(struct lov_user_ost_data_v1);
415 return sizeof(struct lov_user_md_v3) +
416 stripes * sizeof(struct lov_user_ost_data_v1);
419 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
420 * use this. It is unsafe to #define those values in this header as it
421 * is possible the application has already #included <sys/stat.h>. */
422 #ifdef HAVE_LOV_USER_MDS_DATA
423 #define lov_user_mds_data lov_user_mds_data_v1
424 struct lov_user_mds_data_v1 {
425 lstat_t lmd_st; /* MDS stat struct */
426 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
427 } __attribute__((packed));
429 struct lov_user_mds_data_v3 {
430 lstat_t lmd_st; /* MDS stat struct */
431 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
432 } __attribute__((packed));
435 struct lmv_user_mds_data {
436 struct lu_fid lum_fid;
442 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
443 LMV_HASH_TYPE_ALL_CHARS = 1,
444 LMV_HASH_TYPE_FNV_1A_64 = 2,
447 #define LMV_HASH_NAME_ALL_CHARS "all_char"
448 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
450 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
451 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
452 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
453 #define lmv_user_md lmv_user_md_v1
454 struct lmv_user_md_v1 {
455 __u32 lum_magic; /* must be the first field */
456 __u32 lum_stripe_count; /* dirstripe count */
457 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
458 __u32 lum_hash_type; /* Dir stripe policy */
459 __u32 lum_type; /* LMV type: default or normal */
463 char lum_pool_name[LOV_MAXPOOLNAME + 1];
464 struct lmv_user_mds_data lum_objects[0];
465 } __attribute__((packed));
467 static inline int lmv_user_md_size(int stripes, int lmm_magic)
469 return sizeof(struct lmv_user_md) +
470 stripes * sizeof(struct lmv_user_mds_data);
473 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
475 struct ll_recreate_obj {
481 __u64 id; /* holds object id */
482 __u32 generation; /* holds object generation */
483 __u32 f_type; /* holds object type or stripe idx when passing it to
484 * OST for saving into EA. */
492 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
493 const struct obd_uuid *u2)
495 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
498 static inline int obd_uuid_empty(struct obd_uuid *uuid)
500 return uuid->uuid[0] == '\0';
503 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
505 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
506 uuid->uuid[sizeof(*uuid) - 1] = '\0';
509 /* For printf's only, make sure uuid is terminated */
510 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
515 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
516 /* Obviously not safe, but for printfs, no real harm done...
517 we're always null-terminated, even in a race. */
518 static char temp[sizeof(*uuid)];
519 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
520 temp[sizeof(*uuid) - 1] = '\0';
523 return (char *)(uuid->uuid);
526 /* Extract fsname from uuid (or target name) of a target
527 e.g. (myfs-OST0007_UUID -> myfs)
528 see also deuuidify. */
529 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
533 strncpy(buf, uuid, buflen - 1);
534 buf[buflen - 1] = '\0';
535 p = strrchr(buf, '-');
540 /* printf display format
541 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
542 #define FID_NOBRACE_LEN 40
543 #define FID_LEN (FID_NOBRACE_LEN + 2)
544 #define DFID_NOBRACE LPX64":0x%x:0x%x"
545 #define DFID "["DFID_NOBRACE"]"
551 /* scanf input parse format -- strip '[' first.
552 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
553 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
554 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
555 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
557 #define SFID "0x"LPX64i":0x%x:0x%x"
564 /********* Quotas **********/
566 #define LUSTRE_QUOTABLOCK_BITS 10
567 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
569 static inline __u64 lustre_stoqb(size_t space)
571 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
574 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
575 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
576 #define Q_GETOINFO 0x800102 /* get obd quota info */
577 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
578 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
580 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
581 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
582 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
583 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
584 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
585 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
586 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
587 /* lustre-specific control commands */
588 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
589 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
591 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
593 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
596 #define N_PERMS_MAX 64
598 struct perm_downcall_data {
604 struct identity_downcall_data {
611 struct perm_downcall_data idd_perms[N_PERMS_MAX];
615 /* for non-mapped uid/gid */
616 #define NOBODY_UID 99
617 #define NOBODY_GID 99
619 #define INVALID_ID (-1)
628 #ifdef NEED_QUOTA_DEFS
630 #define QIF_BLIMITS 1
632 #define QIF_ILIMITS 4
636 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
637 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
638 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
639 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
642 #endif /* !__KERNEL__ */
644 /* lustre volatile file support
645 * file name header: .^L^S^T^R:volatile"
647 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
648 #define LUSTRE_VOLATILE_HDR_LEN 14
650 typedef enum lustre_quota_version {
652 } lustre_quota_version_t;
654 /* XXX: same as if_dqinfo struct in kernel */
662 /* XXX: same as if_dqblk struct in kernel, plus one padding */
664 __u64 dqb_bhardlimit;
665 __u64 dqb_bsoftlimit;
667 __u64 dqb_ihardlimit;
668 __u64 dqb_isoftlimit;
690 struct obd_dqinfo qc_dqinfo;
691 struct obd_dqblk qc_dqblk;
693 struct obd_uuid obd_uuid;
696 /* swap layout flags */
697 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
698 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
699 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
700 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
701 #define SWAP_LAYOUTS_CLOSE (1 << 4)
703 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
704 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
705 struct lustre_swap_layouts {
714 /********* Changelogs **********/
715 /** Changelog record types */
716 enum changelog_rec_type {
718 CL_CREATE = 1, /* namespace */
719 CL_MKDIR = 2, /* namespace */
720 CL_HARDLINK = 3, /* namespace */
721 CL_SOFTLINK = 4, /* namespace */
722 CL_MKNOD = 5, /* namespace */
723 CL_UNLINK = 6, /* namespace */
724 CL_RMDIR = 7, /* namespace */
725 CL_RENAME = 8, /* namespace */
726 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
727 CL_OPEN = 10, /* not currently used */
728 CL_CLOSE = 11, /* may be written to log only with mtime change */
729 CL_LAYOUT = 12, /* file layout/striping modified */
733 CL_HSM = 16, /* HSM specific events, see flags */
734 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
740 static inline const char *changelog_type2str(int type) {
741 static const char *changelog_str[] = {
742 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
743 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
744 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
747 if (type >= 0 && type < CL_LAST)
748 return changelog_str[type];
752 /* per-record flags */
753 #define CLF_FLAGSHIFT 12
754 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
755 #define CLF_VERMASK (~CLF_FLAGMASK)
756 enum changelog_rec_flags {
757 CLF_VERSION = 0x1000,
760 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID
764 /* Anything under the flagmask may be per-type (if desired) */
765 /* Flags for unlink */
766 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
767 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
768 /* HSM cleaning needed */
769 /* Flags for rename */
770 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
772 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
773 * has an archive in backend */
776 /* 12b used (from high weight to low weight):
781 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
782 #define CLF_HSM_ERR_H 6
783 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
784 #define CLF_HSM_EVENT_H 9
785 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
786 #define CLF_HSM_FLAG_H 11
787 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
788 #define CLF_HSM_SPARE_H 15
789 #define CLF_HSM_LAST 15
791 /* Remove bits higher than _h, then extract the value
792 * between _h and _l by shifting lower weigth to bit 0. */
793 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
794 >> (CLF_HSM_LAST - _h + _l))
796 #define CLF_HSM_SUCCESS 0x00
797 #define CLF_HSM_MAXERROR 0x7E
798 #define CLF_HSM_ERROVERFLOW 0x7F
800 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
802 /* 3 bits field => 8 values allowed */
814 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
816 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
820 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
822 *flags |= (he << CLF_HSM_EVENT_L);
825 static inline __u16 hsm_get_cl_flags(int flags)
827 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
830 static inline void hsm_set_cl_flags(int *flags, int bits)
832 *flags |= (bits << CLF_HSM_FLAG_L);
835 static inline int hsm_get_cl_error(int flags)
837 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
840 static inline void hsm_set_cl_error(int *flags, int error)
842 *flags |= (error << CLF_HSM_ERR_L);
845 enum changelog_send_flag {
846 /* Not yet implemented */
847 CHANGELOG_FLAG_FOLLOW = 0x01,
848 /* Blocking IO makes sense in case of slow user parsing of the records,
849 * but it also prevents us from cleaning up if the records are not
851 CHANGELOG_FLAG_BLOCK = 0x02,
852 /* Pack jobid into the changelog records if available. */
853 CHANGELOG_FLAG_JOBID = 0x04,
856 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
857 changelog_rec_offset(CLF_SUPPORTED))
859 /* 31 usable bytes string + null terminator. */
860 #define LUSTRE_JOBID_SIZE 32
862 /* This is the minimal changelog record. It can contain extensions
863 * such as rename fields or process jobid. Its exact content is described
866 * Extensions are packed in the same order as their corresponding flags.
868 struct changelog_rec {
870 __u16 cr_flags; /**< \a changelog_rec_flags */
871 __u32 cr_type; /**< \a changelog_rec_type */
872 __u64 cr_index; /**< changelog record number */
873 __u64 cr_prev; /**< last index for this target fid */
876 lustre_fid cr_tfid; /**< target fid */
877 __u32 cr_markerflags; /**< CL_MARK flags */
879 lustre_fid cr_pfid; /**< parent fid */
882 /* Changelog extension for RENAME. */
883 struct changelog_ext_rename {
884 lustre_fid cr_sfid; /**< source fid, or zero */
885 lustre_fid cr_spfid; /**< source parent fid, or zero */
888 /* Changelog extension to include JOBID. */
889 struct changelog_ext_jobid {
890 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
894 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf)
896 size_t size = sizeof(struct changelog_rec);
898 if (crf & CLF_RENAME)
899 size += sizeof(struct changelog_ext_rename);
902 size += sizeof(struct changelog_ext_jobid);
907 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
909 return changelog_rec_offset(rec->cr_flags);
912 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
914 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
918 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
920 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
922 return (struct changelog_ext_rename *)((char *)rec +
923 changelog_rec_offset(crf));
926 /* The jobid follows the rename extension, if present */
928 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
930 enum changelog_rec_flags crf = rec->cr_flags &
931 (CLF_VERSION | CLF_RENAME);
933 return (struct changelog_ext_jobid *)((char *)rec +
934 changelog_rec_offset(crf));
937 /* The name follows the rename and jobid extensions, if present */
938 static inline char *changelog_rec_name(const struct changelog_rec *rec)
940 return (char *)rec + changelog_rec_offset(rec->cr_flags &
944 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
946 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
949 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
951 char *cr_name = changelog_rec_name(rec);
953 return cr_name + strlen(cr_name) + 1;
957 * Remap a record to the desired format as specified by the crf flags.
958 * The record must be big enough to contain the final remapped version.
959 * Superfluous extension fields are removed and missing ones are added
960 * and zeroed. The flags of the record are updated accordingly.
962 * The jobid and rename extensions can be added to a record, to match the
963 * format an application expects, typically. In this case, the newly added
964 * fields will be zeroed.
965 * The Jobid field can be removed, to guarantee compatibility with older
966 * clients that don't expect this field in the records they process.
968 * The following assumptions are being made:
969 * - CLF_RENAME will not be removed
970 * - CLF_JOBID will not be added without CLF_RENAME being added too
972 * @param[in,out] rec The record to remap.
973 * @param[in] crf_wanted Flags describing the desired extensions.
975 static inline void changelog_remap_rec(struct changelog_rec *rec,
976 enum changelog_rec_flags crf_wanted)
981 crf_wanted &= CLF_SUPPORTED;
983 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted)
986 /* First move the variable-length name field */
987 memmove((char *)rec + changelog_rec_offset(crf_wanted),
988 changelog_rec_name(rec), rec->cr_namelen);
990 /* Locations of jobid and rename extensions in the remapped record */
991 jid_mov = (char *)rec +
992 changelog_rec_offset(crf_wanted & ~CLF_JOBID);
993 rnm_mov = (char *)rec +
994 changelog_rec_offset(crf_wanted & ~(CLF_JOBID | CLF_RENAME));
996 /* Move the extension fields to the desired positions */
997 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
998 memmove(jid_mov, changelog_rec_jobid(rec),
999 sizeof(struct changelog_ext_jobid));
1001 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1002 memmove(rnm_mov, changelog_rec_rename(rec),
1003 sizeof(struct changelog_ext_rename));
1005 /* Clear newly added fields */
1006 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1007 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1009 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1010 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1012 /* Update the record's flags accordingly */
1013 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1016 struct ioc_changelog {
1023 enum changelog_message_type {
1024 CL_RECORD = 10, /* message is a changelog_rec */
1025 CL_EOF = 11, /* at end of current changelog */
1028 /********* Misc **********/
1030 struct ioc_data_version {
1032 __u64 idv_flags; /* See LL_DV_xxx */
1034 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
1035 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
1038 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1041 #define dot_lustre_name ".lustre"
1044 /********* HSM **********/
1046 /** HSM per-file state
1047 * See HSM_FLAGS below.
1050 HS_EXISTS = 0x00000001,
1051 HS_DIRTY = 0x00000002,
1052 HS_RELEASED = 0x00000004,
1053 HS_ARCHIVED = 0x00000008,
1054 HS_NORELEASE = 0x00000010,
1055 HS_NOARCHIVE = 0x00000020,
1056 HS_LOST = 0x00000040,
1059 /* HSM user-setable flags. */
1060 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1062 /* Other HSM flags. */
1063 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1066 * All HSM-related possible flags that could be applied to a file.
1067 * This should be kept in sync with hsm_states.
1069 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1072 * HSMÂ request progress state
1074 enum hsm_progress_states {
1081 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1084 case HPS_WAITING: return "waiting";
1085 case HPS_RUNNING: return "running";
1086 case HPS_DONE: return "done";
1087 default: return "unknown";
1094 } __attribute__((packed));
1097 * Current HSM states of a Lustre file.
1099 * This structure purpose is to be sent to user-space mainly. It describes the
1100 * current HSM flags and in-progress action.
1102 struct hsm_user_state {
1103 /** Current HSM states, from enum hsm_states. */
1105 __u32 hus_archive_id;
1106 /** The current undergoing action, if there is one */
1107 __u32 hus_in_progress_state;
1108 __u32 hus_in_progress_action;
1109 struct hsm_extent hus_in_progress_location;
1110 char hus_extended_info[];
1113 struct hsm_state_set_ioc {
1114 struct lu_fid hssi_fid;
1116 __u64 hssi_clearmask;
1120 * This structure describes the current in-progress action for a file.
1121 * it is retuned to user space and send over the wire
1123 struct hsm_current_action {
1124 /** The current undergoing action, if there is one */
1125 /* state is one of hsm_progress_states */
1127 /* action is one of hsm_user_action */
1129 struct hsm_extent hca_location;
1132 /***** HSM user requests ******/
1133 /* User-generated (lfs/ioctl) request types */
1134 enum hsm_user_action {
1135 HUA_NONE = 1, /* no action (noop) */
1136 HUA_ARCHIVE = 10, /* copy to hsm */
1137 HUA_RESTORE = 11, /* prestage */
1138 HUA_RELEASE = 12, /* drop ost objects */
1139 HUA_REMOVE = 13, /* remove from archive */
1140 HUA_CANCEL = 14 /* cancel a request */
1143 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1146 case HUA_NONE: return "NOOP";
1147 case HUA_ARCHIVE: return "ARCHIVE";
1148 case HUA_RESTORE: return "RESTORE";
1149 case HUA_RELEASE: return "RELEASE";
1150 case HUA_REMOVE: return "REMOVE";
1151 case HUA_CANCEL: return "CANCEL";
1152 default: return "UNKNOWN";
1157 * List of hr_flags (bit field)
1159 #define HSM_FORCE_ACTION 0x0001
1160 /* used by CT, cannot be set by user */
1161 #define HSM_GHOST_COPY 0x0002
1164 * Contains all the fixed part of struct hsm_user_request.
1167 struct hsm_request {
1168 __u32 hr_action; /* enum hsm_user_action */
1169 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1170 __u64 hr_flags; /* request flags */
1171 __u32 hr_itemcount; /* item count in hur_user_item vector */
1175 struct hsm_user_item {
1177 struct hsm_extent hui_extent;
1178 } __attribute__((packed));
1180 struct hsm_user_request {
1181 struct hsm_request hur_request;
1182 struct hsm_user_item hur_user_item[0];
1183 /* extra data blob at end of struct (after all
1184 * hur_user_items), only use helpers to access it
1186 } __attribute__((packed));
1188 /** Return pointer to data field in a hsm user request */
1189 static inline void *hur_data(struct hsm_user_request *hur)
1191 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1195 * Compute the current length of the provided hsm_user_request. This returns -1
1196 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1198 * return -1 on bounds check error.
1200 static inline ssize_t hur_len(struct hsm_user_request *hur)
1204 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1205 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1206 (__u64)hur->hur_request.hr_itemcount *
1207 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1209 if (size != (ssize_t)size)
1215 /****** HSM RPCs to copytool *****/
1216 /* Message types the copytool may receive */
1217 enum hsm_message_type {
1218 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1221 /* Actions the copytool may be instructed to take for a given action_item */
1222 enum hsm_copytool_action {
1223 HSMA_NONE = 10, /* no action */
1224 HSMA_ARCHIVE = 20, /* arbitrary offset */
1230 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1233 case HSMA_NONE: return "NOOP";
1234 case HSMA_ARCHIVE: return "ARCHIVE";
1235 case HSMA_RESTORE: return "RESTORE";
1236 case HSMA_REMOVE: return "REMOVE";
1237 case HSMA_CANCEL: return "CANCEL";
1238 default: return "UNKNOWN";
1242 /* Copytool item action description */
1243 struct hsm_action_item {
1244 __u32 hai_len; /* valid size of this struct */
1245 __u32 hai_action; /* hsm_copytool_action, but use known size */
1246 lustre_fid hai_fid; /* Lustre FID to operate on */
1247 lustre_fid hai_dfid; /* fid used for data access */
1248 struct hsm_extent hai_extent; /* byte range to operate on */
1249 __u64 hai_cookie; /* action cookie from coordinator */
1250 __u64 hai_gid; /* grouplock id */
1251 char hai_data[0]; /* variable length */
1252 } __attribute__((packed));
1255 * helper function which print in hexa the first bytes of
1257 * \param hai [IN] record to print
1258 * \param buffer [OUT] output buffer
1259 * \param len [IN] max buffer len
1262 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1263 char *buffer, int len)
1265 int i, sz, data_len;
1270 data_len = hai->hai_len - sizeof(*hai);
1271 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1275 cnt = snprintf(ptr, sz, "%.2X",
1276 (unsigned char)hai->hai_data[i]);
1284 /* Copytool action list */
1285 #define HAL_VERSION 1
1286 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1287 struct hsm_action_list {
1289 __u32 hal_count; /* number of hai's to follow */
1290 __u64 hal_compound_id; /* returned by coordinator */
1292 __u32 hal_archive_id; /* which archive backend */
1294 char hal_fsname[0]; /* null-terminated */
1295 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1296 boundaries. See hai_zero */
1297 } __attribute__((packed));
1299 #ifndef HAVE_CFS_SIZE_ROUND
1300 static inline int cfs_size_round (int val)
1302 return (val + 7) & (~0x7);
1304 #define HAVE_CFS_SIZE_ROUND
1307 /* Return pointer to first hai in action list */
1308 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1310 return (struct hsm_action_item *)(hal->hal_fsname +
1311 cfs_size_round(strlen(hal-> \
1315 /* Return pointer to next hai */
1316 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1318 return (struct hsm_action_item *)((char *)hai +
1319 cfs_size_round(hai->hai_len));
1322 /* Return size of an hsm_action_list */
1323 static inline size_t hal_size(struct hsm_action_list *hal)
1327 struct hsm_action_item *hai;
1329 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1330 hai = hai_first(hal);
1331 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1332 sz += cfs_size_round(hai->hai_len);
1338 * describe the attributes to be set on imported file
1340 struct hsm_user_import {
1349 __u32 hui_archive_id;
1352 /* Copytool progress reporting */
1353 #define HP_FLAG_COMPLETED 0x01
1354 #define HP_FLAG_RETRY 0x02
1356 struct hsm_progress {
1359 struct hsm_extent hp_extent;
1361 __u16 hp_errval; /* positive val */
1366 __u64 hc_data_version;
1368 __u16 hc_errval; /* positive val */
1370 struct hsm_action_item hc_hai;
1374 enum llapi_json_types {
1375 LLAPI_JSON_INTEGER = 1,
1381 struct llapi_json_item {
1390 struct llapi_json_item *lji_next;
1393 struct llapi_json_item_list {
1394 int ljil_item_count;
1395 struct llapi_json_item *ljil_items;
1398 /** @} lustreuser */
1400 #endif /* _LUSTRE_USER_H */